Update docs/DifferentiableProgrammingImplementation.md

Co-authored-by: Richard Wei <[email protected]>

Author: philipturner

docs/DifferentiableProgrammingImplementation.md

@@ -107,7 +107,7 @@ The VJP function returns a tuple with two elements:
 
 JVP and VJP functions can be registered using the `@derivative` attribute. JVP functions should return a tuple with labels `(value: ..., differential: ...)`. VJP functions should return a tuple with labels `(value: ..., pullback: ...)`.
 
-For usage examples of the `@derivative` attribute, please see [this mailing list response](https://groups.google.com/a/tensorflow.org/d/msg/swift/k62Uwnc-EYc/Cncb7mxtCAAJ), which shows how to use the `@derivative` attribute to register a derivative for `ax^2 + bx + c`.
+For usage examples of the `@derivative` attribute, refer to the [`@derivative` attribute section](https://github.com/rxwei/swift-evolution/blob/autodiff/proposals/0000-differentiable-programming.md#derivative-attribute) in the proposal.
 
 Note that derivative functions are defined as "JVP/VJP functions taking original arguments", rather than just the "returned differential/pullback" functions. This is because differential/pullback functions may need to refer to intermediate values computed by the original function - this is possible when they are returned closures that capture values.